Quick Start

This guide walks through the OpenAlice loop we recommend trying first: ask for read-only research, let Alice track durable context, turn follow-up work into an issue, and receive the result in the Inbox.

You do not need a broker account to get value from OpenAlice. The core stack is issue + tracked entity + inbox: a Linear-like board for work, an Obsidian-like memory layer for assets and themes, and a delivery channel for finished reports.

Your First Research Task

Open OpenAlice and go to Ask Alice. In the desktop app, this is the first screen after startup. In a source install, open the UI URL printed by pnpm dev (usually localhost:5173).

The Ask Alice composer starts a workspace session behind the scenes: a native coding-agent CLI running in a git-backed directory, with OpenAlice's research and workspace CLIs on PATH.

Ask for a read-only task first:

What can you tell me about AAPL? Give me a quick overview, including the
business, recent performance, valuation, and the main risks.

Alice can pull market data, company profile, key metrics, news, and technical context without any trading account. The result is ordinary workspace work: the agent reads tools, reasons, writes, and can leave durable files behind.

Try a few more:

What are the top gainers in the market today, and which ones look like real
news versus pure momentum?

Calculate the 50-day and 200-day moving averages for NVDA, then explain what
the setup means in plain English.

Search the news for anything about the Fed in the last 24 hours and summarize
what matters for equities.

Track What Should Persist

When Alice finds something you will want to revisit, ask her to track it:

Track AAPL as a stock entity, and create a topic for AI data-center demand.
Link the relevant notes using [[stock-aapl]] and [[ai-data-center-demand]].

Tracked entities are Alice's long-term memory. They create a small cross-workspace index of assets and topics, and [[name]] links let future reports gather backlinks automatically. This is the part that gets better the more you use it: every research note, issue, and report can add context to the same durable graph.

Open the Tracked tab to see the entity list and backlinks. See Entities & Tracking for the full model.

Turn Follow-up Into an Issue

Now make the work self-describing:

Create an issue called "AAPL earnings watch" with medium priority. The goal is
to monitor earnings expectations and update me if the thesis changes.

Alice writes an issue file under .alice/issues/<id>.md. The issue carries board fields like status, priority, and assignee, plus a markdown body with context and links. It shows up on the Issue Board, where you can triage it like a Linear ticket.

Issues are not just reminders. They are work items the agent can read, update, link to tracked entities, and use as context for future runs.

Schedule a Recurring Run

Ask Alice to make the same issue scheduled:

Run this every weekday before the US open. Check AAPL news, price action,
analyst changes, and anything linked to [[ai-data-center-demand]]. If nothing
important changed, stay silent. If something matters, write a short markdown
brief and push it to my Inbox.

Scheduling is not a separate cron UI. Alice adds when, what, and agent to the issue file. The launcher scanner fires due issues as headless workspace runs, and the agent reports back through the Inbox only when useful.

Watch the flow in three places:

Issue Board — the issue shows status, priority, cadence, last-fired, and next-due metadata.
Runs — the headless execution log shows what launched and whether it finished.
Inbox — finished reports arrive as rendered markdown, linked back to the workspace and originating issue.

Read the Report in Inbox

Open the Inbox tab. A good report should be more than a chat reply: it should carry the document Alice wrote, a concise comment, and provenance back to the workspace or scheduled issue that produced it.

This closes the core OpenAlice loop:

research creates context
tracked entities preserve memory
issues describe work and urgency
scheduled runs keep checking
Inbox delivers the artifact when something matters

That loop is the main product surface today. It turns trading into a process an AI agent can manage with you over time.

Optional: Connect a Paper or Demo Trading Account

Trading is available, but treat it as a beta layer after you understand the core loop above. Start with a paper, demo, or testnet account; do not connect live funds until you have validated the workflow yourself.

Open the Web UI's Trading page, add an account through the wizard, pick one of the presets below, and paste the credentials for the matching environment. The wizard tests the connection before it saves; on success, the account becomes a UTA that Alice can research, stage, commit, and submit for your approval.

Preset	Markets	Safer first mode	Current notes
`alpaca`	US stocks and ETFs	Paper Trading	Paper and live use separate API keys. Alpaca is currently the most straightforward securities paper account.
`ibkr-tws`	Stocks, options, futures, FX, bonds	TWS / IB Gateway paper port	Connects through a local TWS or IB Gateway socket, not API keys. More powerful, but setup is heavier.
`longbridge`	HK, US, CN A-shares via Stock Connect, SG equities	Paper Trading	Uses Longbridge OpenAPI credentials. Access tokens expire and must be rotated manually.
Crypto via CCXT	Binance, OKX, Bybit, Bitget, Hyperliquid	Demo Trading or Testnet where available	These presets all use the shared CCXT engine. We have personally tested the listed venues; other CCXT exchanges may work through `ccxt-custom`, but sandbox, demo, symbol, and execution behavior varies by exchange, so expect to validate them yourself.

OpenAlice also starts with keyless read-only crypto data accounts for Binance, OKX, and Bybit, so public market data does not require connecting a trading account. For setup details, caveats, and execution support by broker, see Trading Setup.

Optional: Approval-gated Trade Flow

With a paper account connected, you can ask Alice to prepare a trade:

Buy 1 share of AAPL at market price in my paper account.

Alice will use the UTA beta stack and Trading as Git:

Stage — Alice stages a placeOrder operation
Commit — the operation is committed with a message and hash
Push — you approve in the Web UI before anything reaches the broker

This approval step is always required. Alice should never send orders without your explicit go-ahead.

The UTA / Trading-as-Git stack is still unstable. If you hit errors, confusing broker behavior, or a failed push, bring the error to the community or open a GitHub issue so we can reproduce it.

Next Steps

Issue Board — Track work and schedule recurring runs with markdown issues.
Entities & Tracking — Build long-term memory for assets and topics.
Inbox — See how reports get delivered back to you.
Workspaces — Understand the directory + git + native-agent substrate.
Trading Setup — Optional beta trading setup.

Search Documentation